On Tue, Jan 04, 2022 at 08:36:42AM -0500, Brian Foster wrote: > On Sat, Jan 01, 2022 at 05:39:45PM +0000, Trond Myklebust wrote: > ... > > > > Fair enough. As long as someone is working on a solution, then I'm > > happy. Just a couple of things: > > > > Firstly, we've verified that the cond_resched() in the bio loop does > > suffice to resolve the issue with XFS, which would tend to confirm what > > you're saying above about the underlying issue being the ioend chain > > length. > > > > Secondly, note that we've tested this issue with a variety of older > > kernels, including 4.18.x, 5.1.x and 5.15.x, so please bear in mind > > that it would be useful for any fix to be backward portable through the > > stable mechanism. > > > > I've sent a couple or so different variants of this in the past. The > last I believe was here [1], but still never seemed to go anywhere > (despite having reviews on the first couple patches). That one was > essentially a sequence of adding a cond_resched() call in the iomap code > to address the soft lockup warning followed by capping the ioend size > for latency reasons. Huh. I wonder why I didn't ever merge that? I said I was going to do that for 5.14 and ... never did. ISTR Matthew saying something about wanting to key the decision off of the number of pages/folios we'd have to touch, and then musing about adding QOS metrics, me getting fussy about that, trying to figure out if there was a way to make iomap_finish_page_writeback cheaper, and ... <checks notes> ...and decided that since the folio merge was imminent (HA!) I would merge it after all the dust settled. Add several months of Things I Still Cannot Talk About and now it's 2022. :( Ah, ok, I'll go reply elsewhere in the thread since I think my thinking on all this has evolved somewhat since then. --D > > Brian > > [1] https://lore.kernel.org/linux-xfs/20210517171722.1266878-1-bfoster@xxxxxxxxxx/ > > > > > Thanks, and Happy New Year! > > > > Trond > > > > -- > > Trond Myklebust > > Linux NFS client maintainer, Hammerspace > > trond.myklebust@xxxxxxxxxxxxxxx > > > > >
